System of deep learning neural network in prostate cancer bone metastasis identification based on whole body bone scan images

Abstract

The invention provides a system of deep learning neural network in prostate cancer bone metastasis identification based on whole body bone scan images. A pre-processing module receives an input whole body bone scanning image and processes the whole body bone scanning image; a neural network module detects whether the input whole body bone scanning image is the osseous metastasis of the CAC or not. The neural network module comprises a thoracic cavity part network module and a pelvis part network module. The thoracic cavity part network module is used for establishing a first-stage accelerated regional convolutional neural network, and segmenting a training image of a thoracic cavity part according to a whole body bone scanning image; and training a second-stage accelerated regional convolutional neural network by using the training image, and classifying the focus of cancer cell bone metastasis. The pelvic part network module uses a convolutional neural network and comprises the following steps: establishing a first-stage accelerated region convolutional neural network, and segmenting a training image of a pelvic part according to an input whole body bone scanning image; and training a convolutional neural network by using the training image to classify whether the pelvic part is a bone metastasis image.

Description

technical field [0001] The present invention relates to a prostate cancer bone metastasis identification system, in particular to a prostate cancer bone metastasis identification system based on deep learning of whole-body bone scan images. Background technique [0002] In whole-body bone scan images, physicians must distinguish which part of the hot zone is normal bone formation, which part is caused by prostate cancer cells, and which part is caused by injury, so each image requires a doctor to diagnose and determine Diagnosis of prostate cancer with bone metastases. Among them, prostate cancer bone metastasis is mostly affected by the scapula, ribs, spine, hip joints, limbs and other parts. However, distinguishing whether the hot spot is cancer cell bone metastasis or normal bone cell reaction by human eyes not only requires experienced nuclear medicine physicians to do it, but also requires a long time for the physician to interpret. Therefore, after working for a long ...

AI Technical Summary

Problems solved by technology

However, distinguishing whether the hot spot is cancer cell bone metastasis or normal bone cell reaction by human eyes not only requires experienced nuclear medicine physicians to do it, but also requires a long time for the physician to interpret. Therefore, after working for a long time, it will human misjudgment

At the same time, although there are general rules for doctors to interpret images, there is no absolute standard, so the interpretation results of different doctors may vary due to their experience.

Therefore, the work of diagnostic imaging is very labor-intensive, so once the number of cases increases, it will cost a lot of manpower and time

Images